Public Policy Polling this week released survey results that showed likely 2012 voters in Colorado mostly held congressional Republicans to blame for the unpopular debt deal reached in Washington and that they so far strongly preferred President Obama to any likely Republican rival. Critics of the survey howled that left-leaning PPP had skewed the results by oversampling Democratic voters in the state. Yes, PPP surveyed more Democrats than it did either Republicans or Independents, all pretty much registered to vote in equal numbers in the state, Director Tom Jensen told the Colorado Independent, and that apparent oversampling is driven not by pollster ideological bias but by the self-selecting pattern established by Colorado citizens polled– and, he said, that’s why PPP numbers have been proven highly reliable over the last two elections.
“Party identification and party registration are not the same thing. We are asking how voters define their party affiliation, not what they’re registered as,” Jensen wrote in an email. “In our Colorado polling… we have generally found that more voters in the state identify as Democrats than Republicans, regardless of what their official registration is. Our polls in Colorado have often been criticized as ‘too Democratic’ but last year we were the only polling company to show [Senator] Michael Bennet consistently ahead while most polls had him down over the course of the year and we were proven right on that front. We also had Barack Obama leading by a much wider margin than most other pollsters over the course of the year in 2008 and we were proven right on that front as well.
“We do not weight our polls for party… This is just how the numbers come out after adjusting for gender, race, and age– it’s not based on some predetermined notion of what the party numbers should be.”
It’s worth noting that polling methods differ and that’s by design. Legitimate pollsters tweak methods to increase accuracy. Sometimes they succeed and sometimes they fail in those efforts.
Rasmussen Reports, for example, has been criticized as right biased, but the fact is Rasmussen skews right, or at least skewed right last year, not out of ideological wishful thinking but as an open methodological gambit. Last year, for example, the firm bet that doing so would produce results that would more accurately reflect Tea Party-charged actual voting. Rasmussen was right in some cases but was wrong on the Colorado Senate race. Rasmussen pollsters had GOP candidate Ken Buck running consistently ahead of Bennet, who won the seat in an historic nail biter of an election.
Candidates and campaigns and political parties will for obvious reasons lean toward pollsters that report what they think is good news. But the proof of a polling method’s reliability is in the results it notches. Jensen stands by those delivered by PPP.